"The use of the proper error term is important. The random statement
will
show you the table of expected mean squares and you can find the proper
error terms from there. For a good time, you can use the Keppel and
Zeddik Appendix.
it tells you how to find EMS using a very labor intensive process that
hurts.
Use a random statement in SAS!
proc glm;
class genotype strain;
model y=cov genotype genotype*cov strain(genotype);
random strain(genotype);
test H = genotype E = strain(genotype);
by marker;
Using the following EMS table, you can construct the other hypothesis
tests
for the covariate and genotype * covariate interaction:
Source Type III Expected Mean Square
cov Var(Error) + Q(cov,cov*genotype)
genotype Var(Error) + 6.2515 Var(strain(genotype)) +
Q(genotype)
cov*genotype Var(Error) + Q(cov*genotype)
strain(genotype) Var(Error) + 14.886 Var(strain(genotype))
Is there an equivalent statement in Stata for EMSs???
I asked this once before when I had Stata 7. Now I have Stata 8.
If there is not an equivalent statement to help Stata users, is it on
the list of features to add?
Or am I forced to slog through the Keppel procedure by hand?
--
David C. Airey, Ph.D.
Vanderbilt University
Department of Pharmacology
8148-A Medical Research Building Three
465 21st Avenue South
Nashville, TN 37232-8548
(615) 936-1510 (voice)
(615) 936-3747 (fax) [email protected] http://homepage.mac.com/dairey/vita/
